istec data management forum #3 pat burns dean of csu...

ISTeC Data Management Forum #3

Pat Burns

Dean of CSU Libraries & VP for IT

Friday, May 2, 2014

05/02/2014ISTeC DM Forum3 1

“A dangerous, foreboding, or deathlike influence or vapor.”


“We are drowning in information and starving for knowledge.” – Rutherford D. Rogers


Understand needs for data management

Understand support for data management◦ Local IT infrastructure

◦ Central IT infrastructure: Globus

◦ Libraries’ Digital Repository, DM support

◦ Other, external

Develop an institutional approach/strategy?



Data

Understanding

Knowledge

Information

Wisdom

IT Systems:‘Doing’

Increasing access to data = Increasing research productivity?

Innovation!

Analysis

Physical

Experiment

Numerical

experiment

Big

Data


The 4th Way of Doing Science?

Agencies with > $100m research funding must require federally-funded research to be made publically available, including data sets


Pubs & Data

Journal

Publication

w/ Data Sets

Journal

Subscription

Journal

Access

Research

(Grants)


Faculty

Libraries

IT: Networks, Storage,Access Mgmt.


Working Data Sets

Scholarly Data Sets(linked to

pubs)

✔

Locality

Data sets, by themselves, are not “creative works,” and are therefore not copyrightable◦But, publications and user’s manuals describing them are


“I’d sooner show someone my underwear, than share my data sets with them!”


Unfunded mandate for us!◦Faculty

◦RA’s

◦Students

◦ IT

◦Libraries


NSF is SERIOUS About Data Sharing



Required as of Jan. 18, 2011

Sharing data will speed research and economic development!

ARL Meeting Oct. 2012◦ Myron Gutmann,

Assistant Director, National Science Foundation, Directorate for Social, Behavioral & Economic Sciences



Proposal Review‘Big’

Funding


Proposal Review‘Small’

Funding

Produce

Data Set

Measure

Usage

If High ‘Big’ Funding

Pubs w/ Data

Processed Data &

Data

Representations

Data Collections &

Structured DB’s

Raw Data & Datasets


1. Data within published articles

2. Supplementary files to articles

3. Data referenced fromarticles & held elsewhere

4. Published datasets

5. Data on local disks

Discoverable◦ Crawled by Google and others

Accessible◦ Robust infrastructure, accessibility

Organized/searchable ◦Metadata◦ Catalogued

Persistent◦ Persistent Digital Object Identifier (DOI)◦ Preserved and transcoded as needed


> 25% Failure

Rate of URL’s!!

1. Federal agency systems: NIH PubMed Central◦ NASA only other federal agency deploying a DR?

2. Disciplinary repositories

3. Local repositories: CSU’s DigiTool◦ Data Management Plan templates

◦ Discoverability, accessibility, & preservation

◦ Usage stats

◦ Linked to pubs

4. Local file share files from a data store◦ Granular access control via a common infrastructure, e.g.

Globus


1. Where do I put my scholarly data?◦ Wherever it’s free! Locality? Can I move it around?◦ Repository: persistence/longevity◦ Globus-enabled file share

2. How do I expose my data?◦ PubMed Central, NASA repository◦ Disciplinary repository◦ Local repository (metadata)


Association of Research Libraries (ARL) SHARE initiative: local repositories◦ “Shared Access Research Ecosystem”

◦ For preservation and access of scholarly works and data sets

◦ Not for storage of working files

CSU’s DigiTool Digital Repository (DR)



HPC Centers,Disciplinary Repositories

Internet

NASA?CSU

ACNS GlobusFile Store

DigiTool DRMetadata

Local

RemoteGlobus User

AccessControl

If files stored in (any) digital repository◦ ~OK for metadata and discoverability

If files stored elsewhere◦ Metadata should be stored on a digital repository

for discoverability, accessibility, citability, and persistence; and “point” to the data

If files stored locally◦ Can use Globus to manage access granularly

◦ What about backup and preservation?



DigiTool

Local RAID

Off-campus

RAID PreservationRAID

Internet

2X2X DR Site

Total = 6X

2XMain Site

X = Storage Size

NSF likes:◦ Managed jointly by Librarians and IT Professionals

◦ Connected to utrahigh-speed research network

◦ The best metadata, standards,…

◦ Discoverability (crawled by all services), accessibility

◦ For preservation of scholarly data and pubs

Limitations◦ Very limited access control (NSF likes this too)

◦ Does not support structured data (e.g. databases)

◦ Not for working data sets


1. Cultural change

2. More work – ugh!

3. Greater capital cost◦ The cost of additional storage

required could be extreme


It depends!


SSD’s are required for performance, but expensive

Amazon announced ~80% cost reduction in storage - $10/TB-mo!◦We are exploring!!!



Big Data

Us

If “incidental,” can be included as a direct cost in the grant

Some storage for scholarly data sets “free” to CSU Faculty◦ First 100 GByte free, thereafter $4k/TB

Storage of pubs+ always “free” to CSU faculty and students◦ Assumes no “big” embedded data

Exploring Amazon as an option


Game

Changer!

For a single file◦ Small: 1 GByte

◦ Medium: 10-100 GBytes

◦ Big: >100 GBytes

For a set of files◦ 100X that of a single file?

How many files is ‘big?’


1 full rack stores:~350 Tbytes usable

& Costs ~$150k(~$1,000/TB)


ACNSStorageHPC

& Other Devices

LocalStorage

04/08/14CC-NIE Panel 40

BiSON

FRGPResearchNetworks

Internet

Border 1Border 2

ProductionLAN

Core 2 Core 1

Campus 100 GigCore Routing

Cluster

DYNES Server &Storage, FDT

“ScienceDMZ”

Commodity Users &Researchers (typ.)

DynamicVLANs

3 ea. 10 Gbps Wave

DYNESIDC

10 GbpsResearch

Connections (typ.)

~ 40 or 100 Gig?

1. ✔

5. ~

2. ✔

3. ✔

4. ✖

3. ✔

Later in this forum, you will hear more from our consummate experts


ORCID – the Organizational Researcher Contributor ID

An unique numeric identifier to “Connect Research and Researchers”

Makes publication data gathering much easier

Complements the DOI

See http://orcid.org/


http://orcid.org/


So much of what we call management consists of making it difficult for people to work.

- - Peter Drucker

Make Work as Easy as Possible

Do we need a standing Data Management Committee?


45

Move in any direction, as long as its somewhat positive!

05/02/2014ISTeC DM Forum3


Sometimes it is better to journey hopefully, than to

arrive!!!

istec data management forum #3 pat burns dean of csu...

Documents